Abstract

Agriculture's vulnerability to extreme weather events represents an escalating global challenge, with disasters reducing cereal production by 9-10% globally over recent decades (Lesk et al., 2016, Nature). While methodological advances have enabled national-scale loss estimation revealing USD 3.8 trillion in agricultural losses from 1991-2021 (Rajagopalan et al., 2023), critical gaps persist in translating these frameworks to subnational contexts where decision-making occurs. This limitation extends across Southeast Asia, where events like Typhoon Yagi (September 2024) exposed how national and provincial aggregates hide the commune-level damage patterns authorities need for effective Post-Disaster Needs Assessments and targeted response under Sendai Framework Indicator C-2.

This paper presents an operational system adapting FAO's counterfactual yield estimation methodology to district (Admin2) and commune (Admin3) levels across Vietnam's northern provinces. The approach integrates 15 years of administrative agricultural statistics with multi-source satellite observations to develop a scalable foundation for disaster loss estimation in data-constrained environments. The system addresses fundamental challenges in subnational disaster risk mapping: data scarcity at granular scales, spatial heterogeneity in disaster impacts, and the need for timely information delivery to support actionable policy interventions.

The methodological innovation lies in the strategic integration of earth observation data (including vegetation indices such as NDVI and EVI, and precipitation anomalies from SPI) not merely for validation but as integral components of the counterfactual estimation process. This integration enables the system to function across three temporal scales: historical risk characterization through retrospective analysis (2010-2024), near real time event detection using satellite-derived severity proxies, and rapid post-disaster loss quantification leveraging both observed yields and geospatial indicators.

Initial implementation across 28 districts in Lao Cai, Yen Bai, and Thai Nguyen provinces demonstrates the system's capacity to produce commune-level disaster risk maps for rice, maize, and cassava (Vietnam's critical food security crops). The framework operationalizes null-distribution screening to distinguish disaster-induced losses from natural yield variability, while spatial aggregation techniques address edge effects and administrative boundary discontinuities that complicate subnational analysis. Economic loss monetization incorporates hierarchical price structures, utilizing observed market prices where available and development-index-adjusted proxies for data-sparse communes.

The system contributes to multiple policy objectives: strengthening national monitoring of SDG Indicator 1.5.2 and Sendai Framework Indicator C-2 on agricultural disaster losses; providing Vietnamese authorities with evidence-based tools for targeting resilience investments; and establishing a replicable framework adaptable to other regions facing similar data constraints. Beyond disaster assessment, the approach informs agricultural insurance schemes, climate adaptation strategies, and resource allocation for smallholder support programs.

As climate extremes intensify, closing the subnational information gap becomes imperative for agricultural resilience. This work demonstrates that methodologically rigorous, policy-relevant disaster risk assessment at commune scales is achievable through strategic integration of administrative data, satellite observations, and established loss estimation frameworks, offering a scalable pathway for transforming global methodologies into actionable local intelligence.